Investigating protein flexibility with robotics-inspired approaches

Many proteins involve flexible regions, usually called loops, which play key functional roles for antibody-antigen recognition or modulate enzyme activity, for instance. Due to their conformational variability, the structural investigation of loops is challenging. In this context, computational methods are essential to analyze and complement experimental data. A graphical review on computational approaches to flexible loop modeling has been recently published in the journal Current Research in Structural Biology [1] by researchers of RIS team at LAAS-CNRS. This novel publication format is aimed to communicate information in a visually appealing representation, and to be accessible to a broad scientific audience.

One of the most efficient methods to generate conformational ensemble models of flexible protein loops has been developed by the researchers based on robotics and artificial intelligence algorithms [2]. The method, called MoMA-LoopSampler, is now easily accessible to the scientific community via a web server [3].

The RIS team, in collaboration with the Center for Structural Biology of Montpellier, in France (CBS, CNRS/INSERM/Univ. Montpellier), also investigates an extreme case of protein flexibility, the so-called Intrinsically Disordered Proteins (IDPs). IDPs are fully functional despite their lack of a permanent secondary or tertiary structure, and they exploit their plasticity to perform highly specialized tasks that are complementary to those of their globular counterparts. Moreover, malfunction of disordered proteins can induce severe diseases such as cancer or neurodegeneration. Then, LAAS-CNRS and CBS develop a novel approach coupling biophysical methods (Small-Angle X-ray Scattering and Nuclear Magnetic Resonance) and sampling algorithms to study the structural properties of IDPs [4] which are key to decipher the bases of these functional and dysfunctional processes.

Conformational ensemble model generated with MoMA-LoopSampler for a flexible protein loop
whose structure cannot be determined experimentally

Most IDPs functions involve interactions with other proteins. In a recent joint work [5], researchers at LAAS and CBS have reviewed and classified the diversity of these molecular interaction, explaining how molecular modeling can help understanding such complex processes.

Simplified illustration of different types of interactions involving IDPs.
They are classified depending on the ordered/disordered nature of the interacting regions.
The representation is not aimed to be exhaustive, and combined interaction types do also exist.
See [5] for details.

The work realized by the LAAS-CNRS team allows a better understanding of the relationships between protein sequence, structure and function, but also has an applicative goal. Indeed, loop modeling techniques are used in pharmacology for the design of therapeutic molecules, i.e. molecules that can interact with regions of proteins containing loops; or also for the design of antibodies, which recognize the antigen via loops. The researchers have also collaborated with the Sanofi laboratory on this subject in a CIFRE thesis by Amélie Barozet, PhD student within the RiS team. Moreover, these methodologies can also be applied in the field of biotechnologies for the design of optimized enzymes.

References:

[1] Amélie Barozet, Pablo Chacón, Juan Cortés : Current approaches to flexible loop modeling.
DOI : Current Research in Structural Biology, 2021, 3, pp.187-191.
HAL : https://hal.laas.fr/hal-03318652v1

[2] Amélie Barozet, Kevin Molloy, Marc Vaisset, Thierry Simeon, Juan Cortés.
A Reinforcement-Learning-Based Approach to Enhance Exhaustive Protein Loop Sampling
DOI : Bioinformatics, 2020, 36 (4), pp.1099-1106
HAL : https://hal.laas.fr/hal-02289207v1

[3] Amélie Barozet, Kevin Molloy, Marc Vaisset, Christophe Zanon, Pierre Fauret et al.
MoMA-LoopSampler: A web server to exhaustively sample protein loop conformations
DOI : Bioinformatics, In press
HAL : https://hal.laas.fr/hal-03326493v1

[4] Alejandro Estaña, Nathalie Sibille, Elise Delaforge, Marc Vaisset, Juan Cortés et al.
Realistic Ensemble Models of Intrinsically Disordered Proteins Using a Structure-Encoding Coil Database
DOI : Structure, 2019, 27 (5), pp.381-391.e2.
HAL : https://hal.laas.fr/hal-01954977v1

[5] Ilinka Clerc, Amin Sagar, Alessandro Barducci, Nathalie Sibille, Pau Bernadó et al.
The diversity of molecular interactions involving intrinsically disordered proteins: A molecular modeling perspective
DOI : Computational and Structural Biotechnology Journal, 2021, 19, pp.3817-3828.
HAL : https://hal.laas.fr/hal-03281983v1

Contact:
Juan Cortés, research director at LAAS-CNRS, juan.cortes@laas.fr